WIPL-D GPU Solver

The tool for speeding up the simulation using graphics processing units (GPU).

WIPL-D GPU Solver is a module that enables usage of NVIDIA CUDA-enabled GPUs to significantly decrease EM simulation time in WIPL-D. For this purpose, up to three NVIDIA CUDA-enabled GPUs are used. This module provides GPU acceleration of three phases in EM analysis:

• matrix fill-in,

• matrix inversion and

• near-field calculations.

Acceleration of up to 40 times could be achieved on a single personal computer, while maintaining the accuracy of final results. The greatest speed-up was achieved in the most time-demanding part of EM analysis, matrix inversion.

Besides the fact that calculations are done on GPUs, WIPL-D GPU Solver gives us several important improvements in out-of-core (OoC) algorithm, which is used when analyzing electrically large EM structures. These are:

• more than one hard disk can be used in parallel, which significantly increases speed of hard disk I/O operations,

• I/O operations are done in parallel with GPU calculations, and hence hard disk I/O time (almost) will not influence the overall solving time,

• new GPU accelerated OoC reduced algorithm allows matrix inversion time to be almost halved when solving problems whose system matrix is symmetric.

Acceleration that can be expected, when GPU Solver is used, depends on hardware configuration on which EM simulations are performed, although it mainly depends on used GPU(s). There are a lot of parameters of certain GPU model that affect calculations speed. Most important are number of CUDA cores, RAM size, processor clock, and memory bandwidth. The higher these parameters are the faster is the simulation.Besides GPU(s), hardware components with great impact on computation speed are CPU RAM, hard disks, and CPU.

Accelerations that can be achieved when using GPU(s) on a regular PC, e.g. a PC with Intel i7 processor at 2.8 GHz, 4 cores (8 threads), 8 GB of RAM, 2 hard disks (besides the hard disk on which operating system is situated) with read/write speed of approximately 100 MB/s, one NVIDIA GPU and Windows 7, range from 10 to 22 times in comparison to the simulation time performed on CPU.

A more equipped PC, not exactly the one you would use for everyday work, e.g. a PC with Intel’s i7 processor at 2.8 GHz with 4 cores (8 threads), 24 GB of RAM, 4 hard disks (besides hard disk on which operating system is situated) with read/write speed of approximately 100 MB/s, and up to 3 NVIDIA's GTX 480 graphic cards with 1 536 MB of video RAM, accelerates the solution up to 40 times in comparison with what can be achieved with CPU solution.

Using the new GPU Solver's possibilities, the matrix of a problem with half a million unknowns can be solved in only 42 hours.

For more info on GPU configurations and ways to use the speed-up on your computer, feel free to contact us.